Deep convolutional neural network based anomaly detection for transactive energy systems

Abstract

A computer-implemented method for power grid anomaly detection using a convolutional neural network (CNN) trained to detect anomalies in electricity demand data and electricity supply data includes receiving (i) electricity demand data comprising time series measurements of consumption of electricity by a plurality of consumers, and (ii) electricity supply data comprising time series measurements of availability of electricity by one or more producers. An input matrix is generated that comprises the electricity demand data and the electricity supply data. The CNN is applied to the input matrix to yield a probability of anomaly in the electricity demand data and the electricity supply data. If the probability of anomaly is above a threshold value, an alert message is generated for one or more system operators.

We claim: 
     
       1. A computer-implemented method for power grid anomaly detection using a convolutional neural network (CNN) trained to detect anomalies in electricity demand data and electricity supply data, the method comprising:
 receiving electricity demand data comprising time series measurements of consumption of electricity by a plurality of consumers; 
 receiving electricity supply data comprising time series measurements of availability of electricity by one or more producers; 
 generating an input matrix comprising the electricity demand data and the electricity supply data; 
 applying the CNN to the input matrix to yield a probability of anomaly in the electricity demand data and the electricity supply data; and 
 if the probability of anomaly is above a threshold value, generating an alert message for one or more system operators. 
 
     
     
       2. The method of  claim 1 , further comprising:
 receiving time series records of transactive exchanges between the producers and the consumers for energy purchases, 
 wherein the input matrix to the CNN further comprises the time series records of transactive exchanges between the producers and the consumers for energy purchases. 
 
     
     
       3. The method of  claim 1 , further comprising:
 receiving time series pricing data corresponding to pricing of electricity from the one or more producers, 
 wherein the input matrix to the CNN further comprises the time series pricing data. 
 
     
     
       4. The method of  claim 1 , further comprising:
 receiving weather data indicating weather conditions at locations corresponding to the one or more producers, 
 wherein the input matrix to the CNN further comprises the weather data. 
 
     
     
       5. The method of  claim 1 , further comprising:
 receiving weather data indicating weather conditions at locations corresponding to the one or more consumers, 
 wherein the input matrix to the CNN further comprises the weather data. 
 
     
     
       6. The method of  claim 1 , wherein the consumers are all located in a particular substation. 
     
     
       7. The method of  claim 1 , wherein the consumers each located within a microgrid with at least one of the producers. 
     
     
       8. The method of  claim 1 , wherein the plurality of consumers are located in a plurality of substations. 
     
     
       9. The method of  claim 1 , further comprising:
 in response to the alert message, receiving one or more feedback messages from the one or more system operators; and 
 retraining the CNN based on the one or more feedback messages. 
 
     
     
       10. The method of  claim 1 , wherein receiving the electricity demand data comprises collecting meter data from one or more smart meters corresponding to the consumers. 
     
     
       11. The method of  claim 1 , wherein the CNN further outputs an anomaly type in addition to the probability of anomaly and the alert message comprises the anomaly type. 
     
     
       12. A computer-implemented method for power grid anomaly detection using a convolutional neural network (CNN) trained to detect anomalies in electricity demand data, the method comprising:
 receiving electricity demand data comprising time series measurements of consumption of electricity by a plurality of consumers; 
 selecting a subset of the electricity demand data corresponding to a subset of the consumers located within a geographic area; 
 receiving pricing data indicating price of power for delivery to the geographical area at times corresponding to the time series measurements of the electricity demand data; 
 generating an input matrix comprising the subset of electricity demand data and the pricing data; 
 applying the CNN to the input matrix to yield an indication of an anomaly in the electricity demand data; and 
 generating an alert message for one or more system operators based on the indication of anomaly. 
 
     
     
       13. The method of  claim 12 , further comprising:
 receiving time series records of transactive exchanges between the producers and the consumers for energy purchases, 
 wherein the input matrix to the CNN further comprises the time series records of transactive exchanges between the producers and the consumers for energy purchases. 
 
     
     
       14. The method of  claim 12 , further comprising:
 receiving time series records of power availability within the geographic area, 
 wherein the input matrix to the CNN further comprises the time series records of power availability. 
 
     
     
       15. The method of  claim 12 , further comprising:
 receiving weather data indicating weather conditions at locations corresponding to the geographic area, 
 wherein the input matrix to the CNN further comprises the weather data. 
 
     
     
       16. The method of  claim 12 , wherein the geographic area is selected to span a plurality of substations. 
     
     
       17. The method of  claim 12 , further comprising:
 in response to the alert message, receiving one or more feedback messages from the one or more system operators; and 
 retraining the CNN based on the one or more feedback message. 
 
     
     
       18. The method of  claim 12 , wherein receiving the electricity demand data comprises collecting meter data from one or more smart meters corresponding to the consumers. 
     
     
       19. The method of  claim 12 , wherein the CNN further outputs an anomaly type in addition to the probability of anomaly and the alert message comprises the anomaly type. 
     
     
       20. A system for using detecting convolutional neural network (CNN) trained to detect anomalies in electricity demand data, the system comprising:
 a plurality of smart meters collecting electricity demand data comprising time series measurements of consumption of electricity by a plurality of consumers; 
 a parallel processing platform comprising a:
 a host computer configured to (i) receive electricity supply data comprising time series measurements related to the availability of electricity by one or more producers, (ii) and generate an input matrix comprising the electricity demand data and the electricity supply data; 
 a device computer comprising a plurality of processors configured applying the CNN to the input matrix to yield a probability of anomaly in the electricity demand data and the electricity supply data.